Image analysis – Applications – 3-d or stereo imaging analysis
Reexamination Certificate
2000-06-09
2004-03-30
Patel, Jayanti K. (Department: 2625)
Image analysis
Applications
3-d or stereo imaging analysis
C382S285000, C351S206000, C359S462000, C356S012000, C396S018000
Reexamination Certificate
active
06714672
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates generally to computer-implemented adjustment and matching of stereoscopic images of the eye fundus.
2. Description of Background Art
Retinal photography has long been an important tool for general ophthalmology, but the necessity of using traditional still photography has made analysis and detection of pathologies both time-consuming and subject to human error.
Telemedicine systems such as the “Ophthalmic Imaging Network” developed by Ophthalmic Imaging Systems, Inc., of Sacramento, Calif., are still in the experimental phase. They offer physicians a system for evaluating a patient's retinal status. Using these systems, a single computerized retinal image is captured in the physician's office and is electronically transmitted to a reading center, where it is analyzed, and the results returned to the physician for evaluation. However, this scheme requires experienced and certified readers, and involves a significant degree of error, resulting in the need for images having to be reread.
The automated evaluation of eye fundus topography and other features associated with severe ophthalmology diseases such as diabetic retinopathy and glaucoma could save millions of people from blindness.
Diabetic retinopathy alone is the number two leading cause of blindness in the United States, after macular degeneration, causing 10% of new cases of blindness each year. Diabetes is a common disease affecting about 2% of the population. Of these cases, 10-15% are insulin dependent (type 1) diabetics, and the remainder are non-insulin-dependent (type 2) diabetics. After living with diabetes for 20 years, nearly all patients with type-1 diabetes and over 60% of patients with type-2 diabetes show some degree of retinopathy. A diabetic is 25 times more likely to go blind than is a non-diabetic. Because of this increased risk, diabetics require periodic retinal screening, which should be part of the routine care of all patients, because it can significantly reduce the risk of developing diabetic eye disease.
Another disease of the eye is glaucoma. Almost 80,000 Americans are blind as a result of glaucoma, and another one million are at risk for vision loss and may not even be aware of the risk. It fact, glaucoma is one of the leading causes of preventable blindness in the United States, and the single-most common cause of blindness among African-Americans. Glaucoma is often called the “sneak thief” of sight because the most common type causes no symptoms until vision is already damaged. For that reason, the best way to prevent vision loss from glaucoma is to know its risk factors and to have medical eye examinations at appropriate intervals.
However, with the telemedicine systems of today requiring human reading, it is difficult to achieve regular, accurate, and inexpensive screening of diabetics and people at risk for glaucoma and other diseases.
The most important parameter in retinal examination is fundus topography. For this reason, ophthalmologists prefer to analyze fundus images in stereo. For example, it is impossible to see diabetic macula edema, which is the swelling of the most sensitive area of the retina, without stereo photographs. Cystoid foveal changes at the central focusing area of the retina are also difficult to detect without a stereo view. Changes to the optic nerve due to glaucoma are also hard to observe using just one picture. Using a stereo image pair, however, 3D information can be extracted from the images and used for imaging and measurements of fundus topography.
At present, automated evaluation of fundus topography is performed with scanning laser systems, which use multiple laser scans to render 3D volumes and extract depth information for the fundus features. One of the products available on the market is the TOPSS scanning laser tomography system of Laser Diagnostic Technologies, Inc., of San Diego, Calif. The system is a multiple-purpose tomograph for imaging and measurement of fundus topography. The system can image and measure topographic features and changes of the optic nerve head, macula holes, tumors, and edemas. With respect to digital images, it is able to enhance image visualization, make volumetric measurements, draw topographic profiles and compare images. However, while scanning laser systems are able to provide reliable information about retinal topography, they are very expensive and use narrow laser beams, which may be harmful to the eye. It would be desirable to extract the same information at low cost from regular stereo photographs made with digital fundus cameras. However, this requires extraction of 3D information from 2D photographs. This is made difficult or impossible by differences in illumination between images, and stereoscopic distortions leading to an inability to match points in one image to corresponding points in other images, as further described below.
Usually, stereo photographs of the human eye fundus are taken with one camera shifted by a small distance, illuminating the fundus through the pupil as illustrated in FIG.
1
. The shape of the eye fundus is generally spherical, so the difference in color and brightness between stereo images depends on the position of the camera, which illuminates the fundus through the pupil at different angles. For example,
FIGS. 2
a
and
2
b
show a left and a right image of an ocular nerve, respectively. In the figures, the left part of the left image in
FIG. 2
a
is darker than the left part of the right image in
FIG. 2
b,
and the right part of the right image is darker than the right part of the left image. In order to be able to perform a matching analysis on the images and create a topographical representation of the fundus, these illumination errors must be substantially reduced or eliminated. In addition, it is often desirable to compare two images of the same fundus taken at different times, or with different cameras. This additionally presents a situation where the illumination in each image may be different, requiring correction before appropriate analysis can take place.
It is possible to adjust the brightness, contrast and color of two images using a histogram adjustment method, as proposed by Kanagasingam Yogesan, Robert H. Eikelboom and Chris J. Barry in their paper, “Colour Matching of Serial Retinal Images,” Lions Eye Institute and Centre for Ophthalmology and Visual Science, Perth, Western Australia, Australia, published in “Vision Science and its Applications,”
OSA Technical Digest
(Optical Society of America, Washington D.C., 1999, pp.264-267), which is incorporated by reference herein in its entirety. In their paper, the authors propose a color-matching algorithm that equalizes the mean and standard deviation of each of the three colors in the image. First, the entire image is split into the colors red, green and blue; the mean and standard deviation are calculated, and then the histograms of both images are adjusted to equalize the images. The color image is reconstituted by recombining the three channels. The problem with this method of adjustment is that the equalization adjustment is made for the whole image, so the differences in illumination within the images remain unchanged. For example, consider the points
202
a
and
202
b
in
FIGS. 2
a
and
2
b,
respectively. From the figures, it can be seen that point
202
a
is much darker than point
202
b.
However, since
202
a
and
202
b
actually are the same point on the eye fundus, both points should ideally be illuminated equivalently. Because the Kanagasingram et al. method uses a histogram to adjust the brightness of the whole image, if
FIG. 2
a
were lightened, for example, points
202
a
and
202
b
might end up being equally bright, but point
204
a,
which was originally lighter than point
204
b,
would now be even brighter, causing increased differences in illumination between the points
204
a
and
204
b.
Thus, adjusting the entire image to compensate for different illumination is not a satisfactory solution. W
Berestov Alexander L.
Garland Harry T.
Carter Aaron
Patel Jayanti K.
LandOfFree
Automated stereo fundus evaluation does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Automated stereo fundus evaluation, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Automated stereo fundus evaluation will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3224970